Interestingly none of those numbers is enough to give a significant difference between failure rates in a chi-squared test with p < 0.05 - not even if you pool all the Soyuz variants. Though they do all hit p < 0.10.
Thanks for running the numbers, which I think prove that it's impossible to say that Soyuz is more reliable than Falcon, even if you count AMOS-1 (which feels like it should be counted) and CRS-1 (which I don't).
